Data Augmentation in a Hybrid Approach for Aspect-Based Sentiment Analysis

Data augmentation is a way to increase the diversity of available data by\napplying constrained transformations on the original data. This strategy has\nbeen widely used in image classification but has to the best of our knowledge\nnot yet been used in aspect-based sentiment analysis (ABSA). ABSA is a text\nanalysis technique that determines aspects and their associated sentiment in\nopinionated text. In this paper, we investigate the effect of data augmentation\non a state-of-the-art hybrid approach for aspect-based sentiment analysis\n(HAABSA). We apply modified versions of easy data augmentation (EDA),\nbacktranslation, and word mixup. We evaluate the proposed techniques on the\nSemEval 2015 and SemEval 2016 datasets. The best result is obtained with the\nadjusted version of EDA, which yields a 0.5 percentage point improvement on the\nSemEval 2016 dataset and 1 percentage point increase on the SemEval 2015\ndataset compared to the original HAABSA model.\n